(navigation image)
Home American Libraries | Canadian Libraries | Universal Library | Community Texts | Project Gutenberg | Children's Library | Biodiversity Heritage Library | Additional Collections
Search: Advanced Search
Anonymous User (login or join us)
See other formats

Full text of "A working hypothesis for further studies of frost resistance of concrete."

Research Laboratory 

of the 
Portland Cement Association 

Bulletin 5 

A Working Hypothesis for 

Further Studies of Frost 
Resistance of Concrete 



February, 1945 



Authorized Reprint from Copyrighted 

Journal of the American Concrete Institute 
New Center Building, 7400 Second Boulevard 

Detroit 2, Michigan 

February 1945, Proceedings V. 41, p. 245 



of the 


Vol. 16 No. 4 7400 SECOND BOULEVARD, DETROIT 2, MICHIGAN February 1945 

A Working Hypothesis for Further Studies of Frost 

Resistance of Concrete 


By T. C. POWERS t 

Member American Concrete Institute 


Basic information is given on the freezing of water in concrete. From 
this information and other published material an explanation of the 
mechanism of the action of frost on concrete is developed. The explana- 
tion takes into account such factors as the degree of saturation of the 
concrete, the permeability and strength of the concrete, hydraulic 
pressures generated during freezing, and air-filled cavities. It is sug- 
gested that the hypothesis be made the basis of further laboratory 
studies of the action of frost in concrete. 


Laboratory testing of concrete to predict its ability to resist frost action 
has been carried on for many years. Nevertheless, no generally satis- 
factory testing technique has been developed; different laboratories 
testing the same kind of concrete often obtain different results and even 
in a given laboratory seemingly contradictory results are not uncommon. 
Thus, we still have a problem of developing a suitable testing technique 
and basis for interpretation of the results. 

The writer has no ready-made solution of this problem to offer, but 
from the writings and experiments of others and from fundamental 
studies carried on in this laboratory it is possible to set up a working 
hypothesis which, together with other hypotheses 3 may eventually lead 
to the desired solution.! To the writer a working hypothesis su< h a> this 
is not something to be accepted because it appears reasonable or rejected 
because it conflict- with previously acquired concepts and therefor* 
appears unreasonable. Rather, it should be a logical development of* 
certain major premises. The development should be such as to point 
out various implications of the premises and thereby to suggest experi- 

*Heceive«l )>\ the Institute Dec. 30, li»44. 
t Manager of Basic- Research, Cortland Cement Assn., Chicago 

t-Many of the iitens presented here are elaborations of those given by W. C. Hansen (Office memorandum, 
P. ('. A. Research Laboratory). 



mental procedure* for U ag the premises. The hypothesis pn nted 
iii this j per rests mainly on the premise that the destruction <>f concrete 
I'V in ■ is caused by hydraulic pressure generated by the expansion 
ompanying freezing <»f water rathei than by direct crystal pressure 
deveL d through growth of bodi< - of ice crystals. I f this premise i 
j ;i then < !«"■- "iln i tint . - brought out in the discussion should ■ 

But piously, if the premise is False, the deductions are likewi 
It is ho i that attempts will be made i<« test the premises experi- 
mentally, h attem] - should Lead to new approaches in the study <•! 

thej i . ili. purpose <-i this paper will have been accora- 

I hyrx parti} '-n certain yet unpublished data on tin 

'■ of ha ned portland i nent p U i r ( <j on the propertii and 
1 of tin • p. ble water that the paste ma] contain h is 

et down this information in ordei thai the hy- 

<" md< ood No at mpt will !>«• ma.le here 

'j |j;| ' ' mi hi - made about the si ructure 

'•• - oth* i pap < 'dii: [uentlj . the dis- 

ill iblj a ■ mew h:ii d mat U . 

Struc e ol cement pojte 

I 'i ..I a*- a porous jrcj enveloping illl- 

h>d 1 • colloid I bvd lion i »du< U Drinciualh 

■ ' • i , 

l ii« ■ i y of 1 he p.-i |n i -hi- i wo h i" el - 

'• ' nl itsell i cond (here i- th< 

1 l m - oi i he pores in tin I- 

1 d ; '.ii. - ' him - i |j. diametei -.I a w atei 

W i tog thei with theoth< i olids, 

fill i la ■■..-.. , ,, .,, ,,,,,, ., || J( | ,,. 

bublih tin i m n». i minute eg -.1 

'<•• n I tin o pact • ompl< l\ , 

ll . I. . 1 1n |i:i>1i e pel leated b) i 

11 nil. i. ..mi J . hj Ini i bat are 

UW II) filled , Pfaj, | i 

" ' y« ■ I < u is not • I. at |,m i. ii 

m i Iiat ii.. rjejx ' - ii hl\ in tin ;.n,, i 

d tl m\ J ■- pn. . \ ill 

• ' hU <Ji- - J a 1 lj< oj r ual « 

in is • I.. 

' i 

. I 



1 111 Jlu a1 fj 

, ^' ( i "' f " , " ' , /'\' vl •" "P" N ''- -•« ul HmI-Mi-! JutUu'id 


hydration, studies of which have been reported i a this laboratory ind 
th ' They do no1 readil} refill by absorption oi water from the out- 
side even though the specimen w be stori d under wat 

When n ill particles oi cement pasti in dried, the particl undergo 
permanent, irreversible shrinkage, rhe shrinkage is such thai in 
pa I ' ii inji a water-cement ratio oi aboul 0.5 b] weight th« hanl'-ned 
I- 1 h ran real -<orb only 95 percent of the oluiw of water it orieinalh 
capable oi holding, On the other hand, when jpecimenofi 
tainin i imilai paste is dried, it i- capable of real bing in wa 
than it lo I initial)} (Thi is probabl} due to the n -': ig i ist 

hrinkage offered bj thi a jregate particles I \< ertheless, a Bpeeimi n 
doe* not become inll\ a tu rated l>\ absorption alone, probabh beea 
1 1 m extremclj dense material di iccated duri hydi n is difficult t 
pew I ' a i '• and because i 4 cn1 rapped air 

r l hi pa rtial i mp1 \ ing of i he capillai ii durii h; tl ion I I 

to l" in important factor contributing to what <j oJ tan 

to fro ! a eon< n te ma} ho^ M In- conclusion b i -I on the n nits o 
experiments which howed that when the ai is removed from b dried 
spech i i m fore mi i odui trig w ttei f hus | i mitting I hi w 
the pa t< a pecimen normally able lo wil l 1 00 f 

sing and i ha \\ ing w ill disin teg f e >m plot* in i h 

Tin rjon I nnitj hi i / llml it h 

a bit U i //>/' h> with ' I via / thai 

co tent wa i bi ah that nu alt 

icat oj nth ' t. I hi of roursi i h« m< ■» 1 1 h fi 

' h li\ Kretigi | tnd othi 

As has been found for other porous I uilding n imen ->f 

c( .ii etc will i 'ill com pi on free its dej i 

excet ds some ritii J \ alue that is prol • p 

i sat hi at ion . but, as will I sin n pi g 
that it m be damaged b) I frei ings I 

doe rei it ura i ion 

t K . I K 

o 1 f 

Freezing temperatures 

Th< higl I tern] rat ui w hii h U i 

the degree ol itu tion of thi or nu I um I 

foi ng on 1 1 1 ble w Tl e H 

indii I l*\ the n tivi pi e\ n 

the terete W hen conci is - Jl\ - 

i m only sli i tb I .rl _ ; imoui ii [' 


i ii 

A. K W \ \\t 

i ( . v i r 

r 1 1 ! ! - 


< tfift Hi 





on the concentration of soluble salts in the freezable water. When a 
saturated, hardened paste is cooled sufficiently, water held in the hard- 
ened paste will freeze, but unlike pure water in bulk, not all of it will 
freeze at the same temperature. This, of course, would be the effect 
expected of dissolved salts, but the effect is very much greater than can 
be accounted for by dissolved salts and is in fact a direct effect of the 
colloidal nature of the hydration product-. This effect is such that the 
amount of ice formed in a given saturated specimen is an inverse function 

of the temperature, the lower the temperature the greater the amount 
of ice formed. 

On raising the temperature, the ice disappears progressively and, in a 
saturated specimen, will be completely melted at a temperature slightly 
below 32 F. 

Amount of freezable water 

By a dilatometer method to be described in another paper the amounts 
of freezable water in various saturated pastes have been measured in 
this laboratory. The results can be expressed in the following general 

foi ; 

if, = w t - kw n .... (1) 

where w, = weight of w< sable water ■> a given temperature 

w t = weight of total water in sample, including non-evaporable 

w n = non-evaporable water 

fc == a< tant which varies slightly with the type of cement and 

\ h the freezing temperat ure. 

Th< '-■ rable water, w H} is defined ;i > that which exhibits vapor 

pressure I tha abo I \ x in-- 5 mm ol mercury at 23 C. This is the 

vapor pressure of the systemMg CIO 2H 2 + Mg(C10 4 )s.3H«0 at 23C. 

»«isab it LOpercent more than the amount of "fi* I water" held by a 

ample oven-dried at 105 C under avej gi atmospheric conditions. 

From Eq ttioi l it follows that when w f = 0, w t » kw u . We may 
esignati I 5 particular vaiu o(w t asw t '. For a given cement and freez- 
- mperatu \i thus n to be proportional to the extent of by- 

t ion of . cement 

The proportionalit factor k was evaluated from U m le on five 
different cements. The data were obtained from tmpl en iron, a 

'-up specimens comprising three different mb = 0.3 to 0.6) 

an< ' curing periods (28 i 00 A -. foi each men Th< a ve •» 
value of *wjis obtained lij short i ipolation of a straight-line plot and 
v wind to I L.75. This vaiu< shows tl turated paste to be 

f rom water! le at 15 C, which is virtually all the i l>le 

w otal w ontent n ?t not be moo th • i ; 


the non-evaporable water content. This mean- al><» that the unfreezable 
pari of the evaporable water i> a quantity equal to about :{ 4 of the uon- 
evaporable water content. 

Since the total water contenl of a satuiatetl pa>te is limited by the 
oiiiiin.tl water-cement ratio, tlie above figures indicate thai if the original 
water-cemenl ratio is properly limited the paste will be free from freezable 
water after hydration. The limit will be higher the greater the extenl of 
hydration. The original water ratio is not indicated directly by these 
figures, however, since the total water content includes the main thai 
occur- during the curing period mid is therefore Larger than the original 
water content. For estimating I he amount of I reezable water in terms of 
the original water content, w 0} the following empirical relationship was 
found in hold: 

wj = mi w< - K w (2) 

where n* is an empirical consl ant It- a verage value was found to be 1 1 i> 
K ha the same significance as k in Equation I but isnumeri« dlysmallei 
When Wj = 

u = K ir„ 

where wj is ihc original water contenl of a past* ha ing no freezable 
water after hydrati< >n. The average value i if K w as found to be 1.3. 

'The values of w * can be determined from the extenl of hydration as 
measured by the non-e\ aporable water content. The value so determined 
for cements of average composition was found to be aboul M() - pei g ol 
cemenl for specimens cured 28 days, and about 0.32 g pei u of cemenl 
for those cured 90 days, the average aon-ovaporaUe wat< i i ontents being 
230 and 0.246 g \)w g <<r < ement, respectively. 

These values indicate the maximum original water-cemenl ratios for 
pastes having no freezable water after hydration. Sine< all practical 
concrete is made with higher water-cement ratios than these, it contains, 
or i> eapahle of containing, freezable water. The quantity of freezabli 
water can be estimated from I. quation 2. I i r purposes of estimation, 
tin- may be written 

wj = 1.16 (w e - 1.3 w*) . .;. 

For example, if w c = 0.5, 
then Wf c = 1.16 (0.5 — 1 ,3 > . 

W hen //-, c = 0.2, Wf < = i)2> »■ \n>v of cemenl 

The data given abov< dl pertain to samples cured at normal tempera- 
tures. If a high curing temperature is used, the relationships are much 
different from those jusi described. 

Situation of ice in concrete 

The spaces in undamaged concrete thai may absorb and hold water 
are believed to be as follow-: 


(1) Spaces within the gel sul stance itself; 

(2) The capillary system within the mass of gel substance; 

(3) The capillary system in absorbent aggregate particles. 

Analysis of adsorption isotherms and studies of the density of the 
evaporable water (to he reported in other papers) both indicate that in a 
saturated paste, a quantity of evaporable water 0.75 to 1.0 times the 
quantity of non-evaporable water is "gel water" as distinct from 
"capillary water."* As shown above, independent measurements of 
freezable water showed that the w/<freezable water is an amount equal 
to about 0.75 times the amount of non-evaporable water. This may be 
considered as experimental evidence that the ice formed in the paste 
occurs only in the capillary system within the gel substance. The 
capillaries are thought of as residues of the original water-filled space 
noi filled by gel and hence are probably pocket-like but apparently 

Besides this experimental evidence there are some indirect indications 
that ice is not likely to form within the gel substance. The average 
specific volume of the water in the gel is about 0/85. This indicates that 
all or most of it is within the range of the surface forces of the solid phase 
and that therefore the spaces in which the water is held are exceedingly 
small. In fact, there i- reason to believe that the relationship between 
the gel water and the gel substance is not. much different from solid 

solution. The probability of nuclei formation in water held in such a 
State is much lower than it is in the larger capillary spaces where, when 

the paste is saturated, the water is under no stress. 

Ice probably does not form in the macro- or even microscopic cavities 
of an undamaged specimen. Such cavities, surrounded by absorbent 
paste, can com;, in do water until some force such as that arising from 
expansion on freezing displaces water from the surrounding paste into 
the cavity. 

Besides forming in capillaries within the paste ice may form also in 
the capillary system of absorbent aggregate particles. 


When concrete specimens are frozen and thawed in water, two dis- 
tinct types of disintegration have been observed: (1) The specimens 

may -how little change in weight and appearance but large losses in 
strength and resilience; or (2) they may show progressive loss by crum- 
bling and spalling but relatively little loss in strength and resilience of the 
remaining material These are extremes. Specimen falling in neither 
category will show both crumbling and loss of strength and resilience of 

of *merc!£y able * aU ' r ™ ' JefiDed M that iiavine a vapor P res s^e at 23 C. greater than about 4 r 10-* 



the remaining material. These are results that an acceptable hypothesis 
must explain satisfactorily. 

Generation of hydraulic pressure 

Ordinary ice can exist under pressures up to about 29,000 psi., but of 
course not at 32 F. At 32 F the pressure cannot exceed 1 atmosphere, 
but for each Fahrenheit degree drop in temperature below the normal 
melting point the pressure may be increased about 736 psi., up to the 
maximum mentioned, which is reached at about —4 F.* For example, 
at 13.5 F ice can exist under a pressure of about 10,000 psi.; or, to put 
it another way, a pressure of about 10,000 psi. would be required to pre- 
vent the formation of ice at 13.5 F if conditions are otherwise such that 
the ice could exist at 32 F under a pressure of 1 atmosphere. 

The above remarks give the magnitude of pressures that would be 
exerted on a piston closing a water-filled cylinder if the piston were held 
so as to prevent the expansion that accompanies freezing. It is thus 
indicative of the magnitude of the pressure that might develop in a com- 
pletely saturated, sealed specimen of concrete. 

Experiments show that fully saturated concrete (saturated with the 
aid of a vacuum pump) is not able to withstand the pressures that are 
developed at the low temperatures used in freezing and thawing tests. 
This shows that concrete is not sufficiently extensible to undergo the 
strains that would be necessary to relieve the pressure. Hence, we may 
conclude that since apparently saturated concrete does not ordinarily 
fail completely on first freezing it is not fully saturated, even after pro- 
longed soaking. That is, ordinarily there is enough residual space in the 
concrete to accommodate the expansion that accompanies freezing. 
Nevertheless, under certain conditions some specimens show loss of 
resilience and abnormal expansion on repeated freezing and thawing 
even though they are not fully saturated; that is, they are damaged by 
freezing even when the degree of saturation of the specimen as a whole is 
below the critical value that would cause bursting and complete disin- 
tegration. For a working hypothesis to be satisfactory it must account 
for this fact. The following discussion is an attempt to meet this re- 

Consider a surface of a specimen that has been in contact with water 
tor some time prior to the beginning of the freezing cycle. The water 
content of the concrete near the surface is probably at or near total satu- 
ration, and is higher, though perhaps only slightly, than the average 
water content of the specimen. If this surface is so situated that heat 
from the concrete flows through it to the surrounding water, the sequence 
of events should be as follows: First, the water against the surface will 

*N. E. Dorsey, Properties of Ordinary Water Substar (Reinhold Publishing Corp., 1940), p. 4f>7. 


freeze, thus ruling th urface of the concrete: second, the water in the 
apillary sj ~ of the concrete nearest the surface will freeze and as 

age of - ak place the -till unfrozen water in the ^turated 

region will be dl-pl ed towar te It- 3 saturated interior. This may be 

vis sed with the help of Fie. 1. Th -drawing represent- a crc -ction 

specimen, the -urfa in question being normal to the 
p >f the pa^ aa. A is the saturated region near the surface: B U 

e region « er water »ntent. 

Fig. 1 


1 rm i the o Is s before any will form in the 

Im = - -ft! trees _ oint mentioned abov 

* l P r y ' T ( < water in the c aen to 

Freeni . o- •• in a before it don in B, b. t of th 

du* _ adi and \ A is a1 i hig water 

- B : 1m - tIi. wit.-r in A /-on 


• • : w in B 


- iri n . A. the unfrozen water will be dtt 

ac< wai the le- ra 1 r poo B. Ii water w«re free t 

res di pro.- wl r would develop 

H ■ - Bdl throug! fine-? ired, 

n. if will gi\ . - to a COT- 

* • ai. and gradients of 

686 - ttt of tl r a iing U 


I re; agai: ; ..• iratei inward > Jtl- 

be reg led as § . abl. . llf , t: 

J ioesi n -nitud. of hydraulic pressure n bi 

as* Nibliahed* \\ IX> 

1 u * - - rati* i.o by i igbt, . 

■" 10 i lo- -\* H e, 


Q~40 X l<r V H 

of .m^k. ^^ ^^ ^"^ P* "* ft - <* «■**«• per eeeoad per foot 


where Q = rate of flow through the concrete, ft. per second 

H = pressure-drop through the layer, ft. of water 
L = thickness of layer, ft. 

If we assume a given rate of water movement (corresponding to a 
given rate of freezing), then we can compute the required difference in 
pressure over a given thickness of the region through which the water is 
being forced. However, for the present purpose it is advantageous to 
estimate the rate of movement required to develop a reacting pressure 
sufficient to damage a specimen of concrete. If we assume that the con- 
crete has a tensile strength of 500 psi. and that the mobile water con- 
tacts 10 percent of the cross-section, then it would follow that a hydraulic 
pressure of 5000 psi. would be required to produce an average stress equal 
to the strength of the concrete. However, the specimen would probably 
be damaged at a lower hydraulic pressure since stresses adjacent to the 
source of pressure would be well above the average. It seems safe to 
assume that a pressure of 2500 psi. would be destructive. (Such a pressure 
could be developed at any temperature below about 28.5 F.) 

With this assumption regarding the magnitude of a damaging pressure 
we can compute the velocity with which the water must be forced through 
a given thickness to give rise to reacting hydraulic pressure of destruc- 
tive magnitude. If the layer through which the water moves is 0.001 ft. 
(0.012 in.), then to develop a reacting pressure of 2500 psi. (5800 ft.-head), 
the rate of flow would be 

40X10" 12 X5800 


= 232 millionths of a ft. per sec. 
= 10 inches per hour. 

A rate of flow of 10 inches per hour would correspond to a rate of ice 
formation of about 100 inches per hour. Such a high rate of freezing is 
improbable, even under laboratory conditions; hence, if damage results 
from freezing through the resistance to movement of water, the thick- 
ness of the layer through which the water is forced to move must be 
considerably in excess of that assumed in the above computation, 0.012 
inches. Since the velocity necessary to develop a given reacting pressure 
is inversely proportional to the thickness of the region through which 
the water is being forced, the required rate for different thicknesses can 
be readily seen from the above computation. For example, if the thick- 
ness is 10 times that assumed, 0.12 instead of 0.012, the rate of flow 
would be 1 inch per hour and the rate of freezing about 10 inches per 
hour; or if the thickness were 0.24 inches, the rate of water movement 
would be ]/2 inch per hour and the rate of freezing, 5 inches per hour, 
and so on. 


Depths of saturation in excess of 34 inch and rates of freezing of the 
order last mentioned are well within the range to be expected in many 
laboratory freezing tests. Hence, to consider the reacting hydraulic 
pressure accompanying water movement as a possible source of dis- 
ruptive pressure is in line with experimental data on the tensile strength 
and degree of permeability of concrete. It indicates a possible mech- 
anism whereby a specimen of concrete can be damaged by freezing even 
though it may not be saturated to the critical degree. 

In connection with the frost heaving of soils, Stephen Taber* has 
described another mechanism of pressure development in porous bodies 
that are not fully saturated. In soils that are not too highly impermeable 
or that are not carrying too great a superimposed load or that are not 
cooled too rapidly, pressure is developed by t lie formation of segregated 
bodies of ice crystals which grow in a direction parallel to the heat flow 
by receiving water from the unfrozen interior. This theory will be dis- 
cussed more fully in a lat--r section of this paper. At this point it will 
suffice to >:iy that the writer believes that the same reasoning and ex- 
periments which demonstrated the soundness of Taber's theory with 
respect to the behavior of certain soils leads to the conclusion that the 
same phenomenon is not likely to occur in concrete, at least under the 
conditions of laboratory freezing and thawing tests. Consequently, 

the hypothesis here developed is based on the assumption that hydraulic 

pressure is the primary disruptive force. In Taber's lac ;■<*, the 
assumption is that concrete behaves as a closed system whereas some 
soils behave as open systems; and the thought is here added that in a 

porous body which acts as a closed system disruptive back-pressure may 
be developed by motion of the water even when there is enough room 

within the body to accommodate the total expansion. However, the 
phenomenon described by Taber is probably not entirely absent, as will 
be brought out Later. 

Crumbling and spading 

On repeated freezing and thawing in water, region A should increase 

in thickness according to the amount of water absorbed by the specimen. 
As the thickness of region A increases, the resistance to displacement 

of water out of that region toward the region of lower water Content will 

increase, and when the saturated region becomes sufficiently thick the 

hydraulic pressure will become greater than the strength of the material 

and cause disintegration or spall ing of some part of region A. The 
thickness of region A at the time when disintegration in region A begin* 
is as the critical depth of saturation mentioned before. 

The concept of critical depth of saturation as given above \£ a some- 
what oversimplified picture, for in reality tin saturated region is not 

*J.i r-">(H' 289 (19» ■'■■.-. Bo //. 113 (1930). 


sharply separated from the other; instead, the two regions are joined by 
a transition zone where there is a continuous moisture gradient. More- 
over, the saturated region is seldom if ever completely saturated; even 
after prolonged soaking it contains spaces into which water can be forced 
by high pressure. However, the simplified version will serve the pur- 
pose of this discussion. 

One of the implications of this concept is that if a specimen were 
uniform in structure (homogeneous, in a restricted sense) and not wholly 
saturated at the start, no crumbling or spalling would result from freez- 
ing until a certain amount of water had been absorbed, such as to saturate 
the surface region to the critical depth. 

The critical depth of saturation should be different for different grades 
of concrete because the magnitude of the hydraulic pressure and the 
average stress produced will depend on the following factors: 

(1) The hydraulic pressure will depend on: 

(a) The permeability of the material through which water 

must flow to escape from the saturated region. 

(b) The rate of freezing. 

(c) The amount of water in region A in excess of the crit- 

ical degree of saturation. 

(2) At a given hydraulic pressure the average stress in the concrete 

will depend on the proportion of solids in a unit cross-section. 

The degree of permeability (item (la) above) should have a marked 
effect on the intensity of hydraulic pressure generated in the saturated 
region during freezing. The following illustration will show how this 
comes about: 

The amount of freezable water in saturated paste can be estimated 
from the empirical relationship given before: 

Wf/c = 1.16 (w /c — 1.3 w n /c) 
For an average cement cured 28 days, w?„ may be assumed to be 0.22. 

Wf/c = 1.16 (wjc - 0.29) 
With the aid of this relationship the following table was prepared: 

Properties of the Concrete (Estimated or Assumed) 

Cement content, ib. per cu. yd 

w/c by weight 

Freezable water content at saturation: 

lb. per cu. yd 

cu. ft. per cu. yd 

cu. ft. per cu. ft 

Coefficient of permeability* 

Tensile strength, psi 

*Ruettgers, Vidal, and Wing; ibid. 




150 x 10- 12 


If water is forced through a thickness of 0.001 ft, at the rate of 1 inch 
per hour ( = 23.2 X 10~ 6 ft. per sec), the respective pressures developed 
:i< computed from Darcy's law are: 

Rich concrete 5020 psi 

Lean concrete o7 psi 

These figures arc, of cour-e, significant only on a relative basis, since 
1 1 io rati and thickness are arbitrarily cho-i n figun - However, Bine 
the rate of freezing is probably not appreciably influenced by the quality 

of the concrete, and since a given distance of flow could be found in 
ithei quality, the pressures indicated should be in substantially < rrect 

To illustrate bow the permeability and strength determine the critical 
depth oi sal rati >n, the following computation is Riven, based on an 

! rate of water-movement of 0.1 in. per hour. 

! ! ■ stress in c.-K-h concrete will be 

unit- lulic-pressurei X (fractional area oontaeted In the EQobile watei 
502 X 0.068 = :;ro pi foi the rich oona 

aii'i 6 7 « n HO * riT pei for th< Iran concrete, 

tutation bt _ r liaaed on -a • >\\\ of ^annanon of 00(11 ft. Since, at a given 

n *sue - . | •; 1 1< trial t- lepth of Hat mat ion, the depth of saturation 

vw 1 1 1 # • 1 1 age stress il> 1 1 » nsile strength is 

600 ox 0.001 = 0.0176 n 0.21 in 
•ii and 

300 0.7 X """i o !:•; ft -5.2 in 

1* m 'i]» i 1 i\ thi solute Lues given may be fai from 

i«'i they servi only i«» illustrate the principle on which tin 

hyp -i~ i- I Fori a i n pie, if the actual rate of movement is 1 10 

fa the thickness required for a given average s< rcss is 10 tin 

due j > n h must be borne in mind also that the computations 

imp i the region where ice is forming r- iulh -atu- 

i *> condition is probably m r fulfilled ; hence, the putations 

! I 

The-' f,. \\ti\i m to indicate that ;i rich concrete hould •-how 

more crumbling oi spalling tl n a lean on* ii in thi e th< erif cal 
th of satura m of the rich mix is only one-twenty-fifth that of the 

lean ii look* ai> if the rich mix might show surface disintegration i rlier 

tl • mi on* How to reach such a coi lusion would l>c to ig- 

nore i an n . and to riool the j difference !• en 

tin n ibsorption of such coi tin lean concrete ib>o»l» 

water much n •■ rapidly. Mop r, rhconcref do not Income full} 

■at unit. 1, < oi prol< ged mg unl< capilli action LS augmen- 

'd by tl due n f zing X< slso that < f it t ,k 

iiilarhn. of time for th< tw grades of < bobe< >me«al ited 


to their respective critical depths, the loss per cycle per unit area from 
the leaner concrete would be much greater than that from the richer. 

It is important to note in this connection that under natural conditions 
of exposure, concrete is usually subjected to alternate wetting and drying 
as well as freezing and thawing. If the periods of drying are long relative 
to the period of wetting, dense, impermeable concrete manages to remain 
at a low percentage of saturation as compared with that of low-grade 
concrete which absorbs water quickly. Moreover, impermeability is 
usually the result of using a high cement content and hence such concrete 
reaches a higher degree of desiccation through the contractions that 
accompany hydration.* These advantages, particularly the advantage 
of a low rate of absorption, are not always brought out in laboratory 
tests where small specimens are used. 

Item (lc) above concerning the degree of saturation requires special 
comment. Theoretically, if the water content of region A is about 
90 percent of saturation or less, then all the expansion can be accommo- 
dated without any of the water escaping from A. If the water content 
exceeds 90 percent of saturation, then the excess must escape <»i region 
A will disintegrate; the greater the excess the greater the amount of 
water that must escape in a given length of time; hence, the greater the 
hydraulic pressure developed. 

If in a given specimen the critical depth of saturation happens to be 
equal to or greater than half the thickness of the specimen, then dis- 
integration should not begin until the entire specimen reaches I he critical 
state of saturation. Such a specimen should show little or no disin- 
tegration during the cycles required to bring it to the critical state, 
but when that state is reached the specimen should disintegrate almost 
completely in a few final cycles. This apparently accounts for the 
observed behavior, mentioned earlier, of those specimens that show 
little change in weight or appearance, but, in a final few cycles, rapidly 
lose all strength and resilience. Such specimens are the relatively coarse- 
textured ones of high permeability. 

Damage without disintegration 

The critical depth of saturation discussed above pertains to the depth 
to which the surface region must be saturated before surface disintegra- 
Hon takes place. Extension of the same principles cited before leads to 
the conclusion that concrete may also be damaged even in regions when 
expansion on freezing could be accommodated by available space within 
the region. 

*For a discussion of the importance of this phenomenon see M. A. Swayze, "Early Concrete Volum p 
Changes and Their Control." ACI Journal, Apr. 1042; Proceeding* V. 38, p. 42.">. 


February 1945 

The disruption of porous, water-laden aggregate particles is sometime- 
responsible for this; water in a particle is trapped by ice or by relatively 
impermeable paste surrounding the particle, and the hydraulic pressure 
generated by freezing disrupts the particles and expands the specimen. 
However, a similar result in the absence of such aggregate particles is 
conceivable. Even if the paste is not saturated to the extent where 
pressure causes disruption, the pressure may still be high enough to 
cause some dilation. Such an effect in a low-w/c concrete might develop 
at a lower degree of saturation than in a high-w/c concrete. This is sug- 
gested by the figures given above on relative intensities of hydraulic 
pressure as controlled by permeability. In the example given, it can be 
seen that a condition that would develop negligible pressure in a lean mix 
might develop pressures high enough to cause dilation of a richer mix. 

Some experimental evidence supporting this is given in Fig. 2. Thi 
represents 3x3xlo-in. concrete prisms (0-% in. aggregate) that had been 
cured in the moist room 7 days, stored in air at 50 percent relative humid- 
ity for 6 week-, and then soaked in water 4 days just before the begin- 
ning of the freezing and thawing test. Xote that the richest specimens 
showed relatively little loss in weight but relatively high expansion and 
• hop in Young'- modulus. I'm example, after 100 freezings and thawing:- 
the concrete prisms containing 4 sacks of cement per cubic yard showed 
0.1 per" iii increase in length and those containing 7 sacks showed 0.3 
percenl increase The apparent decreases in Young's modulus were 2^ 
I 65 percent, respectively. However, the surface disintegrations com- 
pare quite differently; the weight loss of the leaner concrete was about 
v percenl and that of the richer, 2 percent, The relative weight Losses 








A ^ 


CJ Gl 

•£ 3 






0! - 

SO Cycles 


Drop in F 

^\ Expansion 


80 ; 

70 o 


60 S 

30 c 

40 > 





10 c 

6789 10 345678 

Cement Content -sac per cubic yard 


Fig. 2— Results of freezing and thawing tests of concrete having 3 to 4 inch slu 



-'■cm to reflect the relative permeabilities and hence rates of absorption, 
as would be expected, but the relative expansion- and losses in resilience 
seem to require the explanation given above. 

These particular specimens contained some unsound chert and thus 
a true explanation may have to include this faetoi also. However, a 
smaller group of tests using wholly sound aggregate gave results similai 
to these. Hence, it seems very doubtful that the small amount of dis- 
ruptive aggregate can account fortheresuil Bhown. Thesi data therefore 

are regarded as showing the effects of dilation "I the paste itself. 

Influence of cavities 

The discussion up to this point has carried the tacit assumption that 
the submici copic pores in the partially desiccated pas t e a re t he only 
spaces available for the movement of water during freezing. If this wer< 
the usual condition, concrete would show ;i very inferior resistance t<> 
weathering. Ordinary concrete contains many air-filled cavities thai ire 
not in direct communication with the surface. Tin-*' cavitii en- 

trained air bubble-, acci ssible p«»ri m the aggregate parti( l< -, and thin 

fissures under the aggregate particl The fissun . formed during the 

period of bleeding, are at first water-filled but may become partially or 
wholly emptied as hydration proceeds Such fissures are more numerous 
and larger the leaner the concn te and the greater the dump ol the ! sh 


All empty cavities of the kind mentioned, especially air bubbli -,mi 

difficult to fill with water. Thej cannot be filled l»y capillai action 
(by soaking) because a liquid cannot spontaneously Bow from a -mall 
capillary to fill a larger one. However, uatei n be forced into a i avity 
by applying external pie-sure or by pumping th< or out the previously 
dried specimen before admitting water. Under normal condition- a 
pressure exceeding one atmosphere is required to fill a cavity, because , 
the cavity fills its contained air becomes compressed. The p 38U] 

enerated during freezing are probably more than suffii nt to force 

water into such spaces. 

It might at first seem that the amount of spac equired to protect 
concrete would depend on the amount of expansion to be a< mmodafc d. 

However, this is not the case, as will now be explained. 

As shown in the example- previously given undei Crumbling and 
Spelling", the freezable water content of saturated concrete would 

ordinarily lie between 0.07 and 0.10 CU. ft. per en. ft. of COicier, , ami 
the expansion of the water on fr zing would be roug ily 0.007 to 0.010 

cu. ft., or 0.7 to 1.0 per cent of the volume of the concrete. Therefore, 
if the concrete contains 0.7 to 1.0 percent air-spaee, the expansion n, 


theoretically, be accommodated. If the space is air-filled at a pressure 
of one atmosphere, and if we assume that the concrete can withstand 
170 atmospheres (2500 psi.) of hydraulic pressure, then we need to pro- 
vide space equal to the volume expansion plus 1/170 of that volume for 
compressed air which is negligible. That is, we need allow space equal 
to only slightly more than 0.7 to 1.0 percent of the volume of the concrete 
to accommodate the expansion. Since concrete ordinarily contains at 
least this much air-filled space, it should be able to withstand freezing 
and thawing almost indefinitely. Such a conclusion is obviously con- 
trary to fact. 

The figures just considered show that if the destructive action of 
freezing is due to hydraulic pressure, the resistance to movement of water 
must be the primary source of pressure, for practically all concrete con- 
tain- enough air-filled space to accommodate the water-to-ice expansion. 
Hence, the distance between air spaces rather than the total volume of 

iir space would be the factor determining the degree of protection, pro- 
vided the total space is at lea-t 1 percent or thereabouts. 

A- will be seen further on in this discussion, the destructive expansion 
could he accounted for by a modification of Taber's hypothesis, that is, 

by assuming that the damage is caused, not by hydraulic pressure, but 
by the growth of submieroscopic bodies of ice analogous to ice lenses. 
This would account nicely for the fact that concrete containing enough 

space to accommodate the ice-to-water expansion may nevertheless be 

damaged on freezing. However, it would fail to provide a very convinc- 
ing explanation for the fact that introducing more air spaces prevents 
"i- greatly reduces damage by frosl action. This is a reason for attempt- 
ing to develop a hypothesis from a premise not involving direct crystal 

]>i -ure. 

To this phase of the hydraulic-pressure hypothesis there is the follow- 
ing important corollary: 

G en a total air spare greater them the possible amount of expansion, 
(he protection of the concrete WtU be greater the smaller (he average size of 

tin Idual air spin . . 

This follow- from the fact that for a given volume of air held within 

prescribed boundaries, the air being divided into a large number of 

parate units, the greater the number of units the clo» r together they 

must be. 

The equally important converse of the corollary is this: 

For a given degree of protection, th< smaller the ntr-fdled cavities the 
mal the total volume of air required. 

Experience -how- that if concrete contains a large number oi small, 
well distributed air cavities, as it di s after it has tw o mixed in the 


presence of a foam stabilizer, its rate of disintegration and rate of internal 
damage can be reduced to a small fraction of the usual rate. In terms 
of the matters discussed above, tin- i- -ecu to he the result <>f providing 
ample room for expansion and of limiting the thickness <>f the permeable 
material ' throne/h which the water tans! movi to find escape from ice pn — nr< 

Other effects may be involved in this case, but they Deed QOl be con- 
sidered in connection with this hypothesis. 

This hypoi he-is provides an explanation for some oi I be results reported 
by Kennedy.* He found thai neal cement prisms containing entrained 

air failed in a freezing and thawing teal more rape II j than did a companion 

specimen containing virtually no air. The published phot< tphs 9Ug- 

gest that unlike air entrained in mortal <>r concrete, the air in tin 

specimens collected in a relatively small aumbei of large cavities -'pi- 
rated by thick layers oi paste As can be -e. n from i lie above discussion, 
such cells would offer relatively little protection. In fact, once the wall 

Of a large I vity LS ruptured, water (oi in t his case < a< b brim • :m tlou 

in and thus transform the cell from a >urce of protection int*. a -out' 
of disruptive pressure. 

Relative permeabilities of paste and aggregate 

If the aggregate particles in a speeinu n of i on< reti ire less permeabl 
than the paste, then the\ should increase the intensity ol hydraulic 
pressure; in the region where i he paste i- saturated, since, I irl of 

the freezable water, they block the most direel path to th( unsaturated 
region. ( >n this basis one would expect concrete containh .. impermeabl) 
iggregate to tend to fail along aggregate mrfa - under the action 
freezing. This has been pointed oul previously by W, < Sans* n 

If the concrete contains unsaturated aggr te particl that a i on 
permeable than the hardened paste, those par l< - should moderate fch 
hydraulic pressure somewhat as cavities do, until tl pari ■><■ 

saturated. When the particle- are saturated, watei must escape into thi 
surrounding paste during freezing or pressures will devek>] thai are I. gh 

enough to disrupt the aggregate particle- and the surrounding material. 

The intensity of the pre->ure in saturated particles thus depends n th< 
permeability of the paste that lies between tin particl.- and the un- 
saturated region. 

Influence of the initiol degree of saturation 

In an earlier part of the discussion it was s d that when concrete is 

partially dried, the initial freezing point or, more urately, thi final 

melting point is lower the greater the d< gree of dryness oi the specimen. 

The final melting point Is, m fact, a function of the relative vapor pi ur 

*Henry I.. Kennedy. "The Function ol Entrained Air in Concrete. \CI Jocrvvi I 

■qs V. {Vi. p ."' , . ., L T V. 

tOffice memorandum. Portland Cemenl \ssociation. Keseanh lj»borator> . 


of the water remaining in the specimen. Thus, the degree of saturation 
not only determines the severity of the effect of freezing, as discussed 
above, but also determines the maximum temperature at which ice 
can exist within the specimen. 

A specimen that has been cured without gain or loss of water, that is, 
one that has been sealed after casting, would show a final melting point 
lower than the normal melting point of the "cement-solution" which it 
contains; the lower the water-cement ratio and the greater the extent of 
hydration the lower would be the final melting point in a specimen cured 
in this manner. Well hydrated, sealed specimens of low water-cement 
ratio probably contain no water that is freezable at temperatures nor- 
mally experienced in this climate. 

When specimens have access to an outside supply, they tend to absorb 
water during the course of hydration, as evidenced by their increase in 
weight. Specimens of high water-cement ratio are able to absorb nearly 
enough water to compensate for the contractions accompanying hydra- 
tion; specimens of ordinary dimensions and low water-cement ratio are 
not able to do so except during the early stages. Thus, the resistance 
of rich specimens to surface disintegration is accounted for by their lesser 
degree of saturation. 

Factors governing the amount of water absorbed during thawing 

It seems fairly evident that the life of a specimen in a freezing and 
thawing test depends on its initial degree of saturation and to a marked 
degree on the rate at which it can absorb water. One of the factors 
involved in this is the mode and rate of thawing. 

If the surface of a frozen specimen is covered with ice, the specimen 

mnot absorb water through that surface until the ice-coating melts. 

After the outside ice has melted, the rate of absorption should depend 

partly on the amount of pressure-deficiency, i.e., the negative pressure, 

within the specimen. This should depend on several factors, as will 

now be shown. 

On freezing a specimen that is in contact with ice, hydraulic pressure 
is created that drives water inward or into cavities adjacent to the frozen 
regions, as already explained. When the temperature of a frozen 
specimen is raised, thawing inside the specimen begins immediately, no 
matter how low the temperature, and the last ice will melt at a temper- 
ature below the melting point of the ice covering the surface. The time 
available for this melting inside the specimen before the outside ice melts 
is believed to be a significant factor. If heat is received rapidly by the 
surface region, most of the outside ice may be melted before ail the ice 
in the specimen disappears, owing to the steepnos of the temperature 
gradient, However, if heat is received slowly, a layer of ice will remain 


on the surface and thus maintain a constant temperature of 32 F for a 
considerable period. 

Under these conditions of thawing, heat flows into the specimen through 
the outside layer of ice. Hence, melting within the specimen will first 
occur near the surface of the specimen and thus produce a layer of melt 
between the outside ice and the still-frozen interior. The contraction 
accompanying melting should create a negative pressure in this layer and 
thus as melting proceeds the water should flow toward this region. 
As long as the outside remains sealed with ice. all such flow must be bad, 
from spaces into which the water was originally displaced. Thus, in this 
circumstance the original distribution of the water tends to be restored 
during the period of thawing. If water from the outside becomes avail- 
able before thawing on the interior is completed, then water should be 
absorbed whether the specimen has become permanently dilated or not. The 
rate of absorption will, of course, depend upon the difference in pressure 
and on the permeability of the material. Specimens of low water-cement 
ratio seem to absorb very little water during the thawing periods. How- 
ever, existing data are un>atisfactory on this point because measurements 
of the weight of the specimen indicate only the net results of any losse- 
due to disintegration and gains due to absorption of water. 

The fact that the melting of the ice in the specimen may precede the 
melting of the outside ice probably is one important factor accounting 
for the resistance to disintegration or scaling of concrete containing an 
air-entraining agent. Because of the sequence of events as described, 
even the air-filled cavities just under the surface tend to be emptied be- 
fore any water from the outside is available to them. If this were not true, 
it would be difficult to account for the resistance to scaling shown even by 
the troweled surface of concrete containing entrained air. 

Freezing and thawing in CaCl 2 brine has been found to be more de- 
structive than freezing and thawing in water. In terms of the hypothesis, 
a specimen frozen in brine should be able to absorb more liquid during 
the thawing period than one frozen in water; hence, the rate of disin- 
tegration in brine should be greater than in water. The greater absorp- 
tion is due to the lower melting point of the brine. A 10 percent CaT'l. 
brine, for example, will be fully melted at about 22 F. Since most of the 
ice within the specimen melts above this temperature, it follows from 
considerations already given that the portion of the thawing period 
during which liquid from the outside can be absorbed is relatively large 
for the specimens frozen in brine. 

Experimental evidence that the amount of absorption per cycle in 
brine is greater than it is in water can be found in data published by 


Hansen.* In Hansen's tests the specimens were given 100 cycles in 
water and then a 10 percent CaCl 2 brine was substituted for the water 
for the rest of the cycles. During the cycles in water many of the speci- 
mens showed very little change in weight after the first 5 cycles. As soon 
as the brine was substituted for the water, the amount of absorption per 
cycle for these specimens showed a marked increase. In this case, the 
specimens that withstood the first 100 cycles in water also stood up well 
in the subsequent cycles in brine, so that the increased rate of absorption 
observed was not clearly connected with an increased rate of disintegra- 
tion. However, the result seems to account at least in part for the gen- 
erally greater severity of the test when specimens are frozen in brine 
instead of water. 

Effect of thickness of outside layer of ice 

It is possible that the thickness of the ice layer formed on the outside 
of the specimen is one of the factors influencing the number of cycles of 
freezing and thawing required to produce a given degree of disintegration. 
This follows from the consideration given above: the thicker the layer of 
ire, the longer will be the period during which the temperature at the 
surface of the specimen remains at 32 F and therefore the more nearly 
will the pressure differences within the specimen produced on melting be 
equalized before outside water becomes available. To produce uniform 
results in a freezing and thawing test in which the specimens are sur- 
rounded with water, it therefore seems important to control carefully the 
amount of ice allowed to form around each specimen. If follows also 


that condition- of thawing should be carefully regulated so that all 
specimens receive heat at the same rate. 

When the practice is followed of freezing the specimens in air and 
1 hawing them in water, a maximum of opportunity for absorption of 
water from an outside source is provided since there would be little if 
ny ice on the surface to keep the specimen sealed while thawing. This 
would be a factor tending to hasten the process of surface disintegration. 
However, the lack of an ice-seal might reduce the effect of freezing, for 
pressures might be relieved by flow to the outside of the specimen as well 

as toward the interior. 

Data published by Hughes j indicate that freezing in air and thawing 
in water is less destructive than freezing in water and thawing in water. 
Thi> would indicate, according to the above paragraph, that the oppor- 
tunity for water to escape to the outside while freezing in air protects 
the specimens more than the free access to water during thawing harms 


*W. C. Hansen. "Influence of >unds. Cements, and Manipulation upon the Resistance of Concrete to 
Freezing and Thawing." ACI Joubn&l, Nov. 1942, Proceeding* V. 39, p. 10 1943). 

tC. A. Hughes, "The Durability of Cement Mori:, the Cement and Method of Testing Major Vari- 
ables." Pror. A.S.T.M.. SS t Part II. 511 (1933). 



Nature of the negative prejiure in concrete 

\\ hen fi /iim orriii- n, ;| i i . ! v .1 .1 m 

1 bal region to anol her, tin i nt.nr m , will 

lowered ab il in p a\ . ii no i | an ioi of il li 

region entered I the di pi > ed ter < oi t a dr * I. 

compri 1 ed For • ample, with :m .m dining wat. 

Ioi CCd Ml' ii Oft <»l 1 I hit li. 1 1: : u tl II 

W Inn I liawiiiM oceui II • • I- I i in tl iwe< 

region below i hat in th< . . .* u. p , | . r • 

k plained before >me ol I pn i <l be d 

eompi - «l t it pi i] 1 1 II p M 

. ,i|»ill ii v for< ' prob >blj foi tfo p 

mi v In linu ui 1 1 How nit: ■ 1 1 1 f > • 

< 'mi iuer a 1 1 M i i ' I region ui - . i m \ 

nil i'i \\ ' a ill i Miin tl .■ v\ liable ii 

(litimis «-i Up i. | Hen th in ioi h II I 

( >7 r< \\ In fi lli I NimIHi t | I 

gion, i hr w i ntent ol n w ill t I) 

\\ hen melt inj cm , I hi 1 1 gion vull 

extent indii atcd I om ol h( la hn 

w tei content would j " u micd I 

prt i HHure \ gh n l*»v\ f tl 

1 1 i mt lit lc* o] • 1 1 nl I n foi drivr *va 

dried ml ill II, ff|| \ f 

i I III :i huh | -Ml tl N 'ill I 1 

MM) p \ I I, i . it r • r • itii II m 4 a r 

pn'ssiif ol 100 perrrh! :in>l thi ng I * 

h:i\ ■inu 1 1 i \ . pn 

of 2«H jit mo hen vvniilt 1 ha \ 1 pn II 

from Mil- the i i If t\ 

ffertn drt i'i m i in \ I 
t ha. win s riod 

Rate of freezing 

A> menl ioi 1 bef< the inl * 

de\ • h»|M .1 durin ■• f reeling of a not w I 

ii 1 1 nf mo\ 

fl Mm , if w ii . rl f i oling I I 


I rate of fn nu; in ai f I 

onditioc uil to t -:•• v\ II • ♦ 

tl sjh Phi' rati in general will dt h • mg 

18 control led hv i amounl kind id rt 



the amount of water surrounding the specimen, etc. The rate of freezing 
at a point in the specimen should depend on the distance of that point 
from the surface, the greater the distance the lower the rate of freezing. 
Since the intensity of t He* generated pressure depends on the rate of 
freezing, it follows that the greater the distance of a point from a surface 
the less the effect of freezing. Consequently, if the rate of freezing is 

ront rolled by the rate of cooling (no supercooling), the size of the speci- 
men could have a considerable influence on the test results; the larger 
the specimen the smaller the internal damage caused by overstraining. 

The Taber-Collins hypothesis 

As mentioned before, Taber advanced a hypothesis to account for the 
expansion of soils that differs materially from that set forth above for 
concrete. Very recently an application of Tabor's theory to the frost 
resistance of concrete lias been published by Collins,* Collins concluded 
from his observations in the field and experiments in the laboratory that 
frost damage in concrete occurs by segregation of ice into layers in the 
oncrete and that the layer- exert pressure by growth in a direction op- 
posite to the How of heal . ( ollins' explanation is as follows: 

Cooling begins at the exposed surface and extends slowly inwards. When any layer 
below the surface reaches a sufficiently Low temperature, the water in the largest pores 
begins to freeze and the latent heat given up by it tends to maintain constant tempera- 
ture at the point of ice formation. The ice crystals so formed are in contact with un- 
frozen water in the surrounding, smaller pores and, by drawing water from them, the 

<•• -tal> continue bo grow. 

The force exerted by iee will be perpendicular to the cold surface; and if 

the concrete is of low strength, a plane of weakness parallel to the cold surface will tend 
i»mi at the level at which the ice is forming. 

The water drawn in by the growing crystals o! ice will come first from (he largest 

unfrozen pores As these become emptied, the supply will l>e restricted and the rate of 

rowth of the ice will be checked. The evolution of latent heat will not then he sufficient 

to maintain the temperature constant at the point of iee-formation and the temperature 

will begin to fall once more. 

As there is then little or no water in the largest pores in the concrete immediate!} 
below the first ice layer, freezing will not begin again until either the temperature has 
dropped sufficiently to freeze the pores tliat do contain water or a level is reached where 
t he larger pores are not affected by the ice forming above them. The result of this 
process ifi that concrete will contain a series of planes of weakness parallel to the surface 
of cooling. During the subsequent cycles of freezing the ice will again tend to form at 
the ime levels as before, because the pores there have been distended by the previous 
ict\ and the freezing point of the water in them will be higher than in the su Wilding 

The damage to tin nc;« e is nsidered to be caused not so much by the actual 
increase in volume of the water in the pores on h /.\n^ as by the growth of the erj ds 
afterwards and the consequent segregation and concentration « i into the layers* 

♦A. H. Collins. J. IruL Cinl /. ._ _ • I'M I 


To substantiate this conclusion Collins made experiments on concrete 
cylinders, of the same kind that Taber made on -oils. In these experi- 
ments the specimen is cooled at one end only while water, kept above the 
freezing temperature, is supplied at the opposite end. When using a very 
low-grade, porous cylinder of conerete, Collins observed the development 
of a horizontal crack about one inch below the top of the concrete cylinder 
and considered this to be analogous to the laminations which appeared 
in a soil cylinder under the same condition of test. 

With respect to the effect of air-entraining agents ( 'ollins -ays: "The 
precise way in which these materials improve durability of concrete is 
not yet fully understood, but they are believed to act in one of two ways. 
They may change the pore structure of the concrete by entraining air or 
by forming bubbles of gas, or they may make the concrete itself water re- 
pellent. In either case the normal movement of moisture within the con- 
crete is disturbed or restricted and the ability of the water to form ic< 

layers is thereby apparently reduecd." 

Collins' observations of the behavior of concrete in the mild climate 
of England led him to the conclusion that ronnvh mixes so propori ioned 
as to have a water-cement ratio of less than 0.7 by weight are able to 
withstand frost action. 


As indicated in the introduction, the writ- r 1 - no desire to try to re- 
fute the Tal )cr-( 'oil ins hypothesis*; on the contrary, he feels thai it should 
be kept in mind and that experiments should he made to test the premise 
on which it is based. If still other hypotheses can be devised, so much 'he 
better, as far as progress toward a correct understanding of this qui stion 
is concerned. However, to justify offering a hypothesis that i< in basic 
disagreement with the ideas sel forth by Collins some discussion <>l* the 
Taber-Collins hypothesis is advisable. 

Although Taber's hypothesis seems highly satisfactory for explaining 
frost heaving in soils, there are several reasons for believing that it should 
nol be applied to concrete or at least not to concrete frozen under the 
usual laboratory conditions. Some of these reasons r t on direct evi- 
dence; some on inference. 

Possibly the most direct evidence against the Taber-< ollins hypotl sis 
is found in a paper published l,y Mattimore.f In this paper -lata are 
given showing that in laboratory tests the h the i e <•!' fi -zing the 
smaller the destructive effect, whereas the Taber-Collins \\\ isia 

requires the opposite result. 

*This designation is used for convenience. The writer does not kno er Taber **r applied 

his hypothesis to concrete or whether he would endorse such an application. 
iProc. Highway Research Board /'/. 135 (1936). 


Inferred evidence against applying the Taber hypothesis to concrete 
comprises those factors and observations indicating that concrete must 
freeze as a "closed system," whereas segregation of ice can occur only in 
"open systems." Taber denned a closed system as "... systems 
from which water could not escape and into which water could not enter,"* 


Leading to this definition is the observation that when a liquid freeze; 
in a porous system, unfrozen liquid at a higher temperature tends to flow 
toward the ice that is first formed and thereby to promote the growth of 
the first-formed ice crystals. When water is available and able to flow 
freely enough to the point of initial ice formation, ice will not form at a 
lower level as long as this condition persists. The condition of flowing 
"freely enough" is one in which the heat given up by the water on freezing 
is enough to maintain constant temperature. 

When the heat is conducted from the region of initial ice formation 
considerably faster than it can be transferred to that region, then the 
water in the specimen freezes in place and lenses do not form. Regardless 
of whether or not outside water is available to enter the specimen, Taber 
calls the condition tn-t described "freezing in an open system," and the 
one next described "freezing m a closed system." 

Taber observed that with respeel to ice segregation, "The chief factors 

are: size and shape of soil particle, amount of water available, size and 

percental of void-, rate of cooling^ and surface load or resistance to 

heaving." These are the factors that determine the permeability of the 

oil, its thermal conductivity, the amount of latent heat released on 

fn i zing, per unit volume of sod, and the mechanical force acting against 
the expansion of the ice. 

Concrete differ- widely horn soil with respect to some of these prop- 
erties Ruettgers, Vidal, and Wingf show the permeability of concrete 
to range from about 16 X 1 <J " 8 ft. per day for concrete having a water- 
cement ratio of 0.44 (5 gal. per sack) to 14,000 x JO- 8 for w c = 1.0 
11 gal. per Back). Dimitri KrynineJ gives the permeability of clay as 
anging from 0,48 ft. per day (48,000,000 X l0- 8 > to "practically zero." 
Pluramei and Dore** give the permeability of clay as "less than 0.00024 
ft. per day" (24,000 X 10- Taber found thai the less permeable clays 
froze closed systems, that is, tin re was no segregation of ice even with- 
out ;i superimposed load. Since Taber diil not publish the permeabilities 

if thes< clays, it i> not certain whether their permeabilities woe "practi- 
cally zero" or somewhat higher. Yet ii is char thai all but the very 
leanest concrete- must have whai the -oil tester would judge by his mei hod 
of test to be practically zero permeability. Amon^ such concretes the 

"J. QsoL SB. 303 [1930). 

: / Mechanic* (McGraw-Hill, 1941) p. 80- 

i Fig, &. 

d J Hummer u 8t i ley M. Dore. S Mt / [Pitman, 1940) p. 45. 


degree of impermeability alone precludes the formation of lenses under 
the usual rates of cooling. Even for the leanest of practical concrete, 
which for want of definite data must be assumed to have permeabilities 
possibly ms high as some clays, we must conclude that lens formation 
is highly improbable because of other factors discussed below. 

When ice tends to expand by lens growl h, it meets much more resistance 
in concrete than it does in unloaded soil. The ice first formed in concrete 
•could occupy not over 10 percent of the cross-section. Tins ice cannot 
grow to form a lens unless it can first disrupt the capillaries in which it 
forms. The force required for disruption will depend on the tensile 
strength of the concrete. As shown earlier, this force could probably be 
as high as 2500 psi. 

In soils the voids run upwards of 35 percent. Then ore, even for the 
same total force the unit pressure on the ice first formed in soil would be 
about J£ to 1/7 of thai in concrete. Ii is very significant that Taber 
found that loading a soil often caused it to freeze as a closed system, 
whereas the same soil formed lenses when frozen unloaded. He says, 

"A relatively small surface load will entirely prevent frost heaving in an 
open system if the material is of such texture that <>nly a little segregat- 
ing ice forms under the most favorable conditions." 

Thus consideration of permeability and resistance to expansion maki s 
it appear that if any kind of sod freezes as a closed system, then all con- 
crete should do so except perhaps when the rate of cooling i- exceedingly 

It would seem reasonable, however, to assume that ice segregation 
takes place in concrete on a submicroscopic scale. As alreadj ointed 
out, 1 he hardened paste is considered to be compost d largely of a porous 
gel containing a system of capillaries. If the ice forms only in the capil- 
laries, as seems likely, then we may imagine [\ to be forming in a container 
having porous, water-laden walls of gel. Hence, the ice c< Id and prob- 
ably would receive water from the walls as well as from the more open 
capillaries. The amount of water available from the ■_ I is vi ry limit I; 
nevertheless some expansive force due to this effect is pro liable. In fai t, 
we could have assumed that the expansion on freezing \- due entirely t«> 
segregation on this submicroscopic scale but, as pointed out before, such 
an assumption does not lend itself readily to an explanation of the pro- 
tective effect of air-filled cavities or the observed effect of increasing the 
rate of freezing. It would require the assumption that te cavities do 
not directly relieve the pressure but merely permit the solid materia] to 
yield under stress with less disruption. How almosl complete immuni 
to frost action could be brought abi r through this mechanism i- not 
very (dear. 




Two corollaries of this hydraulic-pressure hypothesis have been given: 
another will be mentioned here. 

It is apparent that with sound aggregates, and with the concrete not 
initially saturated, the rate of surface disintegration depends mainly on 
the porosity of the paste. The porosity of the concrete as a whole depends 
on the porosity of the paste and on the paste content of the concrete. 
Since paste content is an independent variable, it follows that: 

No general correlation between rate of disintegration and porosity of the 
whole concreU shouldh expected. 

Seeming exceptions to this are the results reported by Hansen* where 
fairly satisfactory correlation between absorption and durability was 
found for cores cut from highways and specimens cast in the laboratory. 
However, the absorptions reported were the quantities of water taken up 
during two day- of soaking, after only two days of drying from a previous- 
ly soaked condition. Owing to the fact that the water lost during a two- 
day period of evaporation can be practically all reabsorbed in two or three 
hours, the value.- reported by Hansen must have depended on the rate 
of evaporation during the drying period. This, in turn, depended not 
primarily on the total e vapor able water content (total porosity) but very 
larg ly on tin 1 permeability of the concrete. Hence, his correlations were 
virtually between permeability and rate of disintegration. Since the 
permeability of such concrete- depend- predominantly on the porosity 
ol the paste, the correlation was virtually on the basis of paste-porosity. 

The bypbtin - -up] >or1 - si rongly the contention of Scholerf and others 
to the em el thai the degree of saturation (or the "saturation coefficient" 
is "of utmi - 1 importance." In fad we may conclude that the two factors 

thai control the lifi oi a specimen in a freezing and thawing tesl are its 
initial saturation coefficient and its permeability, together with these 
same factors for any unsound aggregate in the concrete. The actual life 
of the specimen will depend on the condition.- thai control its absorption 
of water duiing the test, as discussed above, a uming that the specimen 
s a whole wb£ initially below the critical degree of saturation. 


This hypothesis cannol be proved or disproved by data now at hand. 
To te-t the hypothesis, quantitative data on such factor - initial <\<<p-<< 
of saturation ad ab rptivity of the t< 31 specimens must be known 

*W. C. Hansen. "Uniformitj f Corf- ■ Indication of Paverufnt Qualitv." Pr>- 20th Ann. Meeting 
Highway Res. Bo: p. ACIJoubnal. Nov Proc. 105 f 11 it. 

tC H. - ;olei J ity of Concrel Report on Sii Tests -if Concrete and C< rete 

Aggrc 2d Ed p. 2G M 3.T.M. Publication. 


Without such quantitative information, one can use the hypothesis to 
explain any observed result, simply by making appropriate assumptions 
about the unknown quantities. 

For example, more than once it has been observed that prolonging the 
curing period reduces the resistance of laboratory specimens in a freezing 
and thawing test. Such a result can be explained as follows: An increase 
in the amount of hydration has three possible effects : 

(1) It reduces permeability. 

(2) It reduces the amount of freezable water. 

(3) It possibly increases the initial degree of saturation. 

According to the hypothesis a reduction in permeability and an increase 
in degree of saturation are factors increasing the intensity of hydraulic 
pressure and are therefore detrimental to durability. On the other hand, 
the reduction in the amount of freezable water is beneficial. To account 
for the observed effect it is only necessary to assume that the effect of 
reducing permeability or increasing the initial degree of saturation, or 
both, outweighed the effect, of the reduction in freezable water. To 
account for an improvement in resistance from prolonged curing, which 
also has been observed, it is only necessary to make the opposite assump- 

The principal incentive for formulating tins hypothesis is the hope that 
those who have the necessary facilities will begin to accumulate such data 
as are required to support or refute the ideas set forth. This requires 
new knowledge of the properties of paste and of aggregate, particularly 
their relative permeabilities. It also requires data on the permeability 
and absorptivity of the concrete as a whole. It requires new .-Indies 
of supercooling. 

With respect to testing technique it require.- 'lose control of freezing 
and thawing conditions, particularly the rate of cooling, the minimum 
temperature reached, the amount of ice-coating, if any, and the length 
of time during which the surfaces remain at 32 F during the thawing 


It requires an accurate measure of the degree of saturation of each 
specimen at the beginning of the freezing test and a study of the con- 
ditions that control the degree of saturation. 

The hypothesis indicates also that entrained air should lie more 
effective the smaller the bubbles. Hence, if there is any way to control 
bubble-size, it would be advantageous to make the bubbles as small as 
possible and thus minimize the total volume of air required, for to do 
so should hold to a minimum anv adverse effect of entrained air on 


strength. This aspect of the hypothesis calls for laboratory experiments 
along several lines. 

Discussion of this paper should reach the ACI Secretary, in triplicate, 

by April 1 , 1 945, for publication in the JOURNAL for June 1 945. j